I don't think it's not possible to force people to not ask the technical questions. Once it's asked, community decides whether it's on-topic or not. Closing only because it's a technical question isn't enough. More things needs to be taken into the account before deciding.
To be clear, this proposal comes from the Science category, so scientific questions are clearly on-topic (especially socio-scientific angle), but some overlap in scope is expected.
Please note that there are over 10 sites across Stack Exchange network where Artificial Intelligence related questions can be also on-topic (such as Cross Validated, Data Science, Computer Science, CSTheory, Cognitive Sciences, Philosophy, Worldbuilding, Stack Overflow, History of Science, Robotics, GameDev and so on), so once the question is asked, it's a matter of speculation where it exactly should belong, unless it's very clear where it belongs. Otherwise claiming the ownership of some question related to AI on other non-AI site which has been asked specifically here or only because it's a technical one, it would be unwise. The point is, that this site is fully dedicated to AI, Cross Validated site has only few tags related to AI and machine learning and it focuses only on statistical techniques where the questions asked there doesn't have to be related to AI.
Therefore if the question is asking about statistical techniques, then sure, it's more on-topic at Cross Validated. Especially if you think it's off-topic here (e.g. nothing to do with AI), and on-topic there, vote to close, so after the closure it can be migrated by the moderators to another site. Similar with question specifically about data science or programming.
In summary, the level of technicality is a matter of speculation. For me as far as it doesn't consist math, asking for formulas, technical implementation or modelling, programming code, it's not a technical question. We should rather ask ourselves, whether it's off-topic here (non-AI), and on-topic somewhere else.
Related discussion: What should be on-topic, modelling or implementation, or anything else?